SALIENCE ESTIMATION FOR OBJECT-BASED VISUAL ATTENTION MODEL

Description

FIELD OF THE INVENTION

[0001] The present invention relates to a method for estimating the salience of an image, and more particularly to a salience estimation method for object-based visual attention model.

BACKGROUND OF THE INVENTION

[0002] As a neurobiological conception, attention implies the concentration of mental powers upon an object by close or careful observation. Attention area is the area in a picture where tends to catch more human attention. The system designed to automatically detect the attention area of a picture is called attention model. The detected attention area is widely utilized in many kinds of applications, such as accumulating limited resource in an attention area, directing retrieval/search, simplifying analysis, etc.

[0003] Fig.1 indicates the general architecture of a mostly used attention model. First, an image to be estimated is inputted into the attention model. Then the feature of intensity, colour, orientation, etc. will be achieved after the step of feature extraction. In the third step the salience of said features are estimated. After the steps of fusion scheme and post-processing the attention area is finally got.

[0004] Different from attention models used in most previous machine vision systems which drive attention based on the spatial location hypothesis with macro-block (MB) being the basic unit, other models which direct visual attention are object-driven, called object-based visual attention model.

[0005] A lot of researches on MB (macro-block) spatial-based visual attention are established as proposed by L.Itti et al., "A Model of Salience-Based Visual Attention for Rapid Scene Analysis", IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 20, No.11, November 1998 and by Y.F.Ma et al., "A User Attention Model for Video Summarization", ACM Multimedia'02, pp.533-542, December 2002. However, object-based visual attention is not so widely studied because of its inherent difficulty. Y.Sun et al. propose a framework of object-based visual attention in "Object-based Visual Attention for Computer Vision", Artificial Intelligence, pp.77-123, May 2003. Another object-based visual attention model is presented by F.Orabona et al., "Object-based Visual Attention: a Model for a Behaving Robot", 3rd International Workshop on Attention and Performance in Computational Vision, June 2005. Both object-based visual attention schemes still follow the general architecture of attention model listed in Fig.1. All the processes except "salience estimation" are directly inherited from Itti's MB spatial-based visual attention model.

[0006] No matter in MB spatial-based or in object-based visual attention models, low level spatial/temporal features are first extracted, and then for each salient (different, outstanding from its surroundings; or say, more attractive) feature map of each unit is estimated over the whole picture, after that a master "salience map" is generated by feeding all feature maps in a purely bottom-up manner.

[0007] Compared with object-based visual attention model, the MB spatial-based visual attention model is a much easier and faster creation. However, it has several inherent disadvantages:

1) The attention area breaks natural object boundary;

2) Each micro-block may cover lots of natural objects.

[0008] So, the extracted feature of the micro-block is a mixed property of all these natural objects and thus will lower down attention area detection precision.

[0009] The key issue of the object-based visual attention model lies in two aspects: one is the object grouping before feature extraction, the other is the particular efficient salience estimation of each object over all the objects in the image. The central idea of the currently used salience estimation scheme is based on Gauss distance measure as presented by Y.Sun et al.

[0010] Denote x as the object to be salience estimated, y_i(i=1, 2,..., n) as all the background objects, w as the maximum of the width and height of the input image, and //x-y_i// as the physical distance between x and y_i, so the Gauss distance is defined as the formula(1),

with the scale σ set to w/ρ, where ρ is a positive integer and generally 1/ρ may be set to a percentage of w such as 2%, 4%, 5% or 20%, 25%, 50%, etc.

[0011] Denote S_F(x,y_i) as the absolute difference of object x and y_i in feature F, then the salience estimation S_F(x) as the overall salience degree of object x in feature F can be expressed as Formula (2).

[0012] By the definition of the salience estimation, it can be concluded that:

1. The larger difference between the object and its surroundings exists, the more salient the object is.

2. The closer the object and its feature differed surroundings is, the more salient the object is. That is, human vision decreases its ability to distinguish the difference according to distance. The attenuation coefficient is measured by d_gauss, which is coherent with the visual physiology thesis.

[0013] This guarantees S_F(x) is a useful salient estimation in feature F. Unfortunately, some important human perception properties are not considered in S_F(x).

[0014] Fig.2a is an original image of Skating to be estimated and Fig.3a is the salience estimation result of Fig.2a using the conventional object-based visual attention model.

[0015] Fig.2b is an original image of Coastguard to be estimated and Fig.3b is the salience estimation result of Fig.2b using the conventional object-based visual attention model.

[0016] Both in Fig.3a and Fig.3b, white colour means a very outstanding object while black colour means not salient one, the grey level between white and black represents the salience degree.

[0017] From Fig.3a we can see that the audience is considered salient, because its colour greatly differs from its neighbour's, but actually the part of audience contains no details. Viewers usually will not focus on the audience and recognize it as "video texture".

[0018] Also in Fig.3a, there is a little grey block on the left of the female dancer's head. The block consists of a piece of white skating rink which is circled by black male clothing and female skin, and it is salient in this local area. But when all comes to all, this block is a part of the large skating rink and will not attract viewers' attention. This is called "Local effect". Because of the local effect, the accumulated difference between the object and its neighbours is large and thus it is recognized as "salience".

[0019] From forgoing description we can see that the conventional object-based visual attention model is not efficient enough and there are a lot of human vision properties not considered:

1. Object size - The estimation of the influence that the object size on salience degree is a complex problem. For example, (a) if all neighbouring objects y_i are of the same size s and the size of object x decreases from s to 0, as a result the salience degree of x (S_F(x)) will decrease gradually; (b) if all neighbouring objects y_i are of the same size s and the size of object x decreases from s₁ to s₂ (s₁≫s, and s₁>s₂>s), S_F(x) will increase gradually. Thus we know that the relationship between object size and salience degree is not monotonous. And the problem becomes even more complex when each of the objects may have an arbitrary size.

2. Local effect - If an object is not salient among its near neighbours (local area) while the far neighbours are greatly different from the object, there are two possible results: (a) the object is not salient at all inside the whole image; (b) the local area as a whole is salient inside the image with the object being a member of the local area. No matter in which case, the salient degree of the object does not match what defined above.

3. video texture - Suppose the object features of an image are uniformly random, human will usually ignore the details of the whole image and not any object of the image is salient, while the above defined S_F(x) will be a large number for any of the objects in the image.

[0020] With all these limitations, the conventional object-based visual attention model is far from applicable. Therefore an improved object-based visual attention model is desirable.

[0021] LUO J et al. in "On measuring low-level self and relative saliency in photographic images", PATTERN RECOGNITION LETTERS, NORTH HOLLAND PUBL. AMSTERDAM, NL, VOL. 22, NO.2, February 2001, pages 157-169, proposes a method for measurement of relative saliency features, and mentions generally the obtaining of relative saliency features from the feature measures for each feature in each region of the image. Weighting strategies are mentioned for combination of value regions.

[0022] Further, EP1017019A discloses a method for detecting a main object in an image and proposes using a probabilistic reasoning engine to integrate saliency features into estimation of a region of arbitrary shape and size as a main subject of the image.

SUMMARY OF THE INVENTION

[0023] The present invention provides a salience estimation scheme for object-based visual attention model employing a multi-level concentric circled scheme capable of lowering the computing complexity and being more applicable.

[0024] In one aspect, the invention provides a method for estimating the salience of an image according to claim 1. It comprises steps of segmenting the image into a plurality of objects to be estimated; extracting feature maps for each segmented object; calculating the saliences of each segmented object in a set of circles defined around a centre pixel of the object based on the extracted feature maps; and integrating the saliences of each segmented object in the all circles in order to achieve an overall salience estimation for each segmented object. According to one preferred embodiment, the step of extracting feature maps is based on the measure of image colour variation. According to another preferred embodiment, the step of calculating the salience of each segmented object comprises a sub-step of comparing colour features of the object to be estimated with that of any other object in each circle defined around the object to be estimated.

[0025] Advantageously, the object-based visual attention model based on multi-level concentric circled salience estimation scheme of the present invention presents an efficient framework to construct object-based visual attention model, which is of low computing complexity and much more human vision inosculated.

[0026] Other characteristics and advantages of the invention will be apparent through the description of a non-limiting embodiment of the invention, which will be illustrated with the help of the accompanying drawings.

BRIEF DESCRIPTION OF DRAWINGS

[0027] 

Fig.1 illustrates a general architecture of a mostly used attention model;

Fig.2a illustrates an original image of Skating to be salience estimated;

Fig.2b illustrates an original image of Coastguard to be salience estimated;

Fig.3a is the salience estimation result of Fig.2a using the conventional object-based visual attention model;

Fig.3b is the salience estimation result of Fig.2b using the conventional object-based visual attention model;

Fig.4 illustrates the multi-level concentric circled scheme of the salience estimation according to a preferred embodiment of the present invention;

Fig.5 illustrates an example definition of texture(.) in the invention;

Fig.6a is an example of segmentation result of Fig.2a according to the preferred embodiment of the present invention;

Fig.6b is another example of segmentation result of Fig.2b according to the preferred embodiment of the present invention;

Fig.7a illustrates the estimated salience result of Fig.2a using the salience estimation scheme according to the preferred embodiment of the present invention; and

Fig.7b illustrates the estimated salience result of Fig.2b using the salience estimation scheme according to a preferred embodiment of the present invention.

DETAIL DESCRIPTION OF PREFERRED EMBODIMENTS

[0028] The technical features of the present invention will be described further with reference to the embodiments. The embodiments are only preferable examples without limiting to the present invention. It will be well understood by the following detail description in conjunction with the accompanying drawings.

[0029] From the foregoing description we can see that the salience estimation process can be denoted as:

Input: An image I={p_i| i=1...w*h} with width w and height h;

Output: Salience map sal[1...w*h] with sal[i] the salient degree of pixel p_i in the image.

[0030] The method of the present invention mainly includes three steps described as below:

Step 1 - Pre-processing (image segmentation)

[0031] An image I is first decomposed into a set of objects I={o₁, o₂, ..., o_n} in this step. Image-based segmentation and grouping play a powerful role in human vision perception, a lot of researches have been developed in this area. In this invention we adopt the object segmentation scheme proposed by P.F.Felzenszwalb et al., "Image Segmentation Using Local Variation", IEEE Computer Society on Computer Vision and Pattern Recognition, Jun.1998, which is based on measures of image colour variation. Here gives a simple description of this scheme.

[0032] Before processing, an undirected graph H=(V, E) is defined based on the image I, where each pixel p_i of I has a corresponding vertex v_i ∈ V, and an edge (v_i, v_j) ∈ E connects vertices v_i and v_j. The precise definition of which pixels are connected by edges in E depends on the expression (1-1).

For some given distance d., a weight function on the edges, weight(.), provides some non-negative measures of similarity (or difference) between the individual vertices v_i and v_j. Define weight(.) as expression (1-2),

where colour(v_i) is the colour of pixel p_i in the image.

[0033] S={C_i} denotes a segmentation of V and each C_i corresponds to a segmented object. Define the internal variation of C as formula (1-3),

where MST(C, E) is a minimum spanning tree of C with respect to the set of E.

[0034] Define the external variation of two objects C₁ and C₂ as formula (1-4) .

[0035] The process of the segmentation is to make the expression (1-5) satisfied for any two of the segmented objects:

where k is a constant number set to 100 in our implementation. Denote Int(C)+k/|C | as the extended internal variation inside object C.

[0036] To achieve the segmented result, E is first sorted into π=(e₁, e₂, ..., e_m) by non-decreasing edge weight and initially segment the image into w*h single pixel objects, then for each e_q=(v_i,v_j) (q=1, 2, ..., m) repeat the next process: if v_i and v_j belong to different objects and weight (v_i, v_j) is not larger than the extended internal variation(Int(C)+k/|C|) of the two objects they belong to, the two objects are merged to form a new single object.

[0037] It can be seen, this gives an efficient object segmentation scheme which will not cost too much of computing resource. In implementation, here uses an 8-connected neighbourhood for constructing E, that is d=1. Fig.6a and Fig.6b provide the segmentation results of Fig.2a and Fig.2b respectively.

Step 2 - Pre-processing (feature extraction)

[0038] With y_i, r_i, g_i, b_i denoting the luminance, red, green and blue channels of pixel p_i, following we extract the features of each object segmented in step 1.

[0039] Considering the definition of extended internal variation Int(C)+kl|C|, wherein k/|C | is an addition to the internal variation because underestimating the internal variation is bad for preventing components from growing. As a result, small objects are more likely to grow regardless of the internal variation inside. For example, in the Skating example of Fig.6a, the largest part of the black clothing mainly consists of black pixels but still there are some exceptions of white pixels.

[0040] To solve the feature extraction problem in the mentioned situation, an operator Major(f, o) is defined in the feature map F=f(v_i) of an object o={v₁, v₂, ..., v_t}. The returned value of Major (f, o) is the representative feature of the object o which is defined to satisfy (d₁, d₂ and η are constant number, set to 2, 64 and 95% respectively in our implementation):

(1) If there exists a range [min, max] meeting the expression max-min+1≤d₁ and the percentage of features whose values inside the range [min, max] over the whole feature map F={f(v₁), f(v₂), ..., f(v_t)} are not smaller than η, Major(f, o) is then defined as the average value of those features whose values are inside the range [min, max].

(2) Otherwise, if an object size is larger than a constant d₂, the object is divided into two sub-objects by the same process as in step 1 and then loop above step for each sub-object; otherwise if the object is too small, just define Major(f, o) as the average value of all the features.

[0041] With operator Major(f, o), Y_i, R_i, G_i, B_i and Ye_i being defined as the luminance, red, green, blue and yellow channels of object o_i (negative values are set to 0) :









The intensity feature is extracted as formula (2-1).

The "colour double-opponent" based colour features are extracted as formula (2-2) and (2-3).

[0042] Orientation will be a certain complex feature in object based visual attention model. Since all the objects are segmented according to colour variations, the object itself then will not contain any orientation information except the border of the object. Because of this special property of the segmented objects, we will not consider orientation in the implementation.

[0043] Comparing with orientation, motion will be a more possible additional feature since currently optical flow techniques become more and more mature.

[0044] But for simplicity, we only consider the three feature maps I_i, RG_i and BY_i in present invention.

Step 3 - Salience Estimation

[0045] After the above two steps the image I is segmented into objects I=(o₁, o₂, ..., o_n), and three feature maps I_i, RG_i and BY_i (i=1...n) are extracted. The remaining problem is how to estimate the salience map for each feature map F (F∈{I, RG, BY}), denoted as Sal_F(o_i).

[0046] For any object o_i of the image, denote s_i as the size (the number of pixels inside the object) and c_i=(X_i, Y_i) as the centre pixel of the object. X_i and Y_i are described as formula (3-1).

[0047] During the salience estimation process, each pixel of the o_i is indistinctively considered equal to the center pixel c_i, so the object is considered duplicated s_i copies of the center pixel as shown in Fig.4.

[0048] Based on this assumption, there presents a multi-level concentric circled scheme for salience estimation of o_i. In the first step of this scheme, there defines a set of concentric circles circled around the center pixel c_i of the object, C₁...C_t(C_j is an ellipse with horizontal radius r_x and vertical radius r_y, and is called Level j circle) are distributed from the near neighbouring areas of the center pixel c_i to the far neighbouring areas. For each level j circle, estimate the salience of o_i inside c_j, denoted as SalC_F(o_i, C_j), and the overall estimated salience of o_i is then defined as formula (3-2), where k_t is a constant number for linear integration.

[0049] Then, given an area C_j and an object o_i in C_j with feature F extracted over C_j, how to estimate the salience of o_i inside C_j considering human vision properties? Here we first give the definition of the operation SalC_F:

(1) Set S as the set of objects with center pixel inside C_j.

(2) For each object o_u in S, define F'_u=abs(F_u-F_i). Then use the follow formula to calculate the weighted average of F'_u in S.

(3) Define ρ as the percentage of pixels in S with F'_u not larger than avgF':

where bool(exp) returns 1 when exp is a true determinant else returns 0.

(4) With the definition of a detection function texture (.) as shown in Fig.5, SalC_F is set as below,

Where texture(.) is an empirical function of ρ for detection of "audience area", i.e. the area with random featured objects such as audience, which is more expected not to be recognized as attention. The detection function texture (ρ) satisfies that the lower the value of ρ is, the bigger the value of texture (ρ) will be, and thus the more chance this area is recognized as an "audience area" i.e. the video texture of the image. By using this detection function texture (.) there will be lower probability that the non-attention objects in the area are recognized as attention.

[0050] From the description above we can conclude the salience estimation scheme as below:

(a) For each object o_i of the image, define a set of concentric circles C_j (j=1...t).

(b) Calculate SalC_F(o_i, C_j) according to above definition of SalC_F.

(c) Integrate the salience estimation for all C_j according to expression (3-2) to get the overall estimated salience.

[0051] Fig.7a and Fig.7b respectively present the salience estimation experimental results of Fig.2a and Fig.2b by using the salience estimation scheme according to the present invention. By using the present salience estimation method, the audience in Fig.2a and the background in Fig.2b are considered not salient, and the little block on the left of the female dancer's head in Fig.3a is also removed from Fig.7a. The present invention is capable of handing the local effect and video texture and it is more applicable.

[0052] Whilst there has been described in the forgoing description preferred embodiments and aspects of the present invention, it will be understood by those skilled in the art that many variations in details of design or construction may be made without departing from the present invention. The present invention extends to all features disclosed both individually, and in all possible permutations and combinations.

[0053] The present object-based visual attention model based on multi-level concentric circled salience estimation scheme gives a more accuracy on understanding of the image and a far more computing efficiency, it has several advantages as below:

1. The invention presents an efficient framework to construct object-based visual attention model. It is of low computing complexity.

2. The presented framework is much more human vision inosculated. The un-considered human vision properties in conventional schemes (such as object size, local effect and video texture) are well issued.

3. The framework is extendable.

Claims

1. A method for estimating the salience of an image, the salience being a measure for how much salient objects differ from background objects, the method comprising the steps of:

1) segmenting the image into a plurality of separate objects; for each object,

2) extracting a feature value for the object as a function of the pixels of the object,
wherein the feature value is at least one of luminance intensity and colour;

characterized in that the method further includes steps of:

3) selecting the object's neighboring area, wherein the neighboring area is elliptic;

4) weighting the average difference between the feature value of the object and that of each of other objects in the neighboring area by using the size of each of said other objects to get a weighted average feature difference of the object;

5) estimating the salience of the image as a function of the weighted average feature differences of the plurality of objects.

2. The method according to claim 1, wherein when extracting the feature value of an object, if the number of plurality of pixels in the object whose feature values fall into a range, exceeds a predetermined threshold compared with the total number of pixels in the object, the average of the feature values of the plurality of pixels is extracted as the feature value of the object.

3. The method according to claim 1 or 2, wherein for the object, multiple neighboring areas can be selected in step 3), and step 4) is performed in each neighboring area to get a total weighted average feature difference by linearly summing up all the weighted average feature difference of the object in each neighboring area.

4. The method according to one of the claims 1-3, wherein if there are multiple neighboring areas, these multiple neighboring areas are concentric.

5. The method according to claim 4, wherein each neighboring area of an object is centered by the center pixel of the object.

6. The method according to claim 1, wherein in the weighting step each pixel of an object is considered equal to the center pixel of the object.

7. The method according to one of the claims 1-6, wherein the feature value is the value or variation of color, or intensity of the pixels in the object.

8. The method according to one of the claims 1-7, wherein the size of each of other objects is the number of pixels therein.

9. The method according to one of the claims 1-8, wherein an additional feature value is motion.

10. The method according to one of the claims 1-9, wherein the elliptic area has a horizontal radius and a different vertical radius.

Ansprüche

1. Verfahren zum Schätzen der Salienz eines Bilds, wobei die Salienz ein Maß dafür ist, wie viele hervortretende Objekte sich von Hintergrundobjekten unterscheiden, wobei das Verfahren die folgenden Schritte umfasst:

1) Segmentieren des Bilds in mehrere getrennte Objekte; für jedes Objekt

2) Extrahieren eines Merkmalswerts für das Objekt als eine Funktion der Pixel des Objekts, wobei der Merkmalswert die Farbdichteintensität und/oder die Farbe ist;

dadurch gekennzeichnet, dass das Verfahren ferner die folgenden Schritte enthält:

3) Auswählen des Nachbarbereichs des Objekts, wobei der Nachbarbereich elliptisch ist;

4) Gewichten der mittleren Differenz zwischen dem Merkmalswert des Objekts und dem jedes von anderen Objekten in dem Nachbarbereich unter Verwendung der Größe jedes der anderen Objekte, um eine gewichtete mittlere Merkmalsdifferenz des Objekts zu erhalten;

5) Schätzen der Salienz des Bilds als Funktion der gewichteten mittleren Merkmalsdifferenzen der mehreren Objekte.

2. Verfahren nach Anspruch 1, wobei, wenn der Merkmalswert eines Objekts extrahiert wird, der Mittelwert der Merkmalswerte der mehreren Pixel als der Merkmalswert des Objekts extrahiert wird, falls die Anzahl mehrerer Pixel in dem Objekt, deren Merkmalswerte in einem Bereich liegen, im Vergleich zu der Gesamtanzahl von Pixeln in dem Objekt einen vorgegebenen Schwellenwert übersteigt.

3. Verfahren nach Anspruch 1 oder 2, wobei für das Objekt in dem Schritt 3) mehrere Nachbarbereiche ausgewählt werden können, und wobei der Schritt 4) in jedem Nachbarbereich ausgeführt wird, um durch lineares Aufsummieren der gesamten gewichteten mittleren Merkmalsdifferenz des Objekts in jedem Nachbarbereich eine gewichtete mittlere Gesamtmerkmalsdifferenz zu erhalten.

4. Verfahren nach einem der Ansprüche 1-3, wobei diese mehreren Nachbarbereiche konzentrisch sind, falls es mehrere Nachbarbereiche gibt.

5. Verfahren nach Anspruch 4, wobei jeder Nachbarbereich eines Objekts durch das mittlere Pixel des Objekts zentriert ist.

6. Verfahren nach Anspruch 1, wobei in dem Schritt des Gewichtens jedes Pixel eines Objekts als gleich dem mittleren Pixel des Objekts betrachtet wird.

7. Verfahren nach einem der Ansprüche 1-6, wobei der Merkmalswert der Wert oder die Änderung der Farbe oder der Intensität der Pixel in dem Objekt ist.

8. Verfahren nach einem der Ansprüche 1-7, wobei die Größe jedes der anderen Objekte die Anzahl der Pixel darin ist.

9. Verfahren nach einem der Ansprüche 1-8, wobei ein zusätzlicher Merkmalswert die Bewegung ist.

10. Verfahren nach einem der Ansprüche 1-9, wobei der elliptische Bereich einen horizontalen Radius und einen anderen, vertikalen Radius aufweist.

Revendications

1. Procédé d'estimation de la saillance d'une image, la saillance permettant de mesurer à quel point des objets saillants diffèrent des objets d'arrière-plan, le procédé comprenant les étapes suivantes :

1) segmentation de l'image en une pluralité d'objets distincts ;
pour chaque objet,

2) extraction d'une valeur de caractéristique pour l'objet en fonction des pixels de l'objet, où la valeur de caractéristique est au moins l'intensité de luminance ou la couleur ;

caractérisé en ce que le procédé inclut en outre les étapes suivantes :

3) sélection de la zone avoisinante de l'objet, où la zone avoisinante est elliptique ;

4) pondération de la différence moyenne entre la valeur de caractéristique de l'objet et celle de chacun des autres objets de la zone avoisinante en utilisant la taille de chacun desdits autres objets pour obtenir une différence de caractéristique moyenne pondérée de l'objet ;

5) estimation de la saillance de l'image en fonction des différences de caractéristique moyennes pondérées de la pluralité d'objets.

2. Procédé selon la revendication 1, dans lequel lors de l'extraction de la valeur de caractéristique d'un objet, si le nombre de la pluralité de pixels de l'objet dont les valeurs de caractéristique se trouvent dans une plage dépasse un seuil prédéterminé par rapport au nombre total de pixels de l'objet, la moyenne des valeurs de caractéristique de la pluralité de pixels est extraite en tant que valeur caractéristique de l'objet.

3. Procédé selon la revendication 1 ou 2, dans lequel pour l'objet, plusieurs zones avoisinantes peuvent être sélectionnées à l'étape 3) et l'étape 4) est effectuée dans chaque zone avoisinante pour obtenir une différence de caractéristique moyenne pondérée globale par l'addition linéaire de toutes les différences de caractéristique moyennes pondérées de l'objet dans chaque zone avoisinante.

4. Procédé selon l'une des revendications 1 à 3, dans lequel s'il existe plusieurs zones avoisinantes, ces zones avoisinantes multiples sont concentriques.

5. Procédé selon la revendication 4, dans lequel chaque zone avoisinante d'un objet est centré par le pixel central de l'objet.

6. Procédé selon la revendication 1, dans lequel lors de l'étape de pondération, chaque pixel d'un objet est considéré comme égal au pixel central de l'objet.

7. Procédé selon l'une des revendications 1 à 6, dans lequel la valeur de caractéristique est la valeur ou la variation de couleur ou d'intensité des pixels de l'objet.

8. Procédé selon l'une des revendications 1 à 7, dans lequel la taille de chacun des autres objets est le nombre de pixels qu'il contient.

9. Procédé selon l'une des revendications 1 à 8, dans lequel une valeur de caractéristique supplémentaire est le mouvement.

10. Procédé selon l'une des revendications 1 à 9, dans lequel la zone elliptique présente un rayon horizontal et un rayon vertical différent.

Drawing